Christopher Sims, VAR Model And Impulse-Response Analysis

2011 Nobel Economics Prize winner Christopher Sims (who passed away yesterday) revolutionised macroeconomics by developing the Vector Auto Regression (VAR) model and impulse-response analysis.

Till the 1970s, macroeconomic analysis involved taking a large system of equations built around a Keynesian macroeconomic model and statistically estimating them. The estimated system was then used to:

1. Interpret time series

2. Make economic forecasts

3. Conduct policy experiments

These large models were successful in explaining the historical data. But in the 1970s, Western countries had high inflation along with low growth and high unemployment. During this period of 'stagflation', the large models started showing instabilities - which created problems.

In 1980, Christopher Sims wrote a paper titled "Macroeconomics And Reality" (Econometrica, 22k+ citations) to solve this problem. He developed a new methodology using a new model called Vector Auto Regression (VAR). VAR is basically a system of N equations of N variables in which each variable is a function of:

1. Its own past values

2. Past values of the other variables

3. Some exogenous shocks.

His methodology has 3 steps:

1. Estimate the VAR model and forecast the macroeconomic variables - by separating unexpected movements in the variables from expected movements.

2. Identification - ie, break down these unexpected movements into structural shocks (ie, shocks that are the fundamental causes of macroeconomic fluctuations).

3. Impulse-response analysis - ie, trace out the dynamic impact of these shocks on subsequent movements in all the variables.

Sims' new methodology revolutionised macroeconomics. Today, VAR model and impulse-response analysis have become the basic tools of macroeconomic analysis. They are used to answer questions like:

1. What causes economic cycles?

2. What are the effects of fiscal policy?

3. What are the effects of monetary policy?

The pic is from Sims' paper. It shows the impulse-response analysis of America's GDP - ie, it shows how GDP is affected by 6 variables: money supply, GDP itself, unemployment, wages, prices and import prices (top to bottom). As you can see, the graphs do not show the confidence intervals. In the 1970s/80s, computers were not very powerful - so it was difficult to calculate the confidence intervals. In recent times, of course, every impulse-response graph shows the confidence intervals . . .

Info-source: Nobel Prize Org